Oncofusion

start / graph / hyp-menin-inhibition-spares-endogenous-h3k79-unit

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q3b-dot1l-h3k79

Menin inhibitors should leave half the H3K79 methylation standing - and if they do not, the menin-interface model is wrong somewhere

also called two-unit H3K79 hypothesis · menin spares the MLLT10-directed unit · residual H3K79me2 after menin inhibition

A screen found that the H3K79 tag is put down by two separate machines: one steered by the leukaemia fusion and needing menin, and one the normal cell runs through a protein called AF10. If that is right, a menin drug should wipe out the first and leave the second - and measuring exactly how much tag is left is a direct test of the project's central idea.

The same thing, for a scientist

**Statement.** In KMT2A-rearranged leukaemia, menin inhibition depletes the fusion-directed (oncogenic) H3K79 methylation unit and leaves the MLLT10-directed endogenous unit intact; residual H3K79me2 after menin inhibition is therefore a **predicted quantity with a predicted genetic dependency**, not a sign of incomplete response. The hypothesis is a direct extension of CRISPR-ChIP, which partitioned H3K79 methylation into exactly these two units [E1]. **Why it is worth testing rather than assuming.** This project's central bet is the menin interface, and the partition is the strongest unbiased functional-genomic support that bet has received - an assumption-free screen recovering menin as a requirement for the oncogenic half of the mark the disease runs on. It is also the sharpest available **falsifier**: if menin inhibition in a KMT2A-rearranged line removes *all* H3K79me2, or removes *none*, the two-unit model is wrong, and with it the account of why menin inhibitors work. The partner-specificity problem bounds it: **KMT2A::MLLT10 and PICALM::MLLT10 carry AF10 as the fusion partner**, so the two units cannot be separate there [E2], and **KMT2A::AFF1 has neither an AHD nor AF10**, so the route by which the commonest infant fusion reaches DOT1L is not established at all. Prior probability 0.55.

Go deeper
The predictions, in the order they should be tested
  1. Residual H3K79me2 after menin inhibition is MLLT10-dependent. Treat a KMT2A::MLLT3 or KMT2A::MLLT1 line with a menin inhibitor to full target engagement, measure H3K79me2 by calibrated ChIP, then knock out MLLT10 and repeat. The residual fraction should collapse. If MLLT10 knockout does nothing to the residue, the endogenous unit is directed by something else and the partition is mis-assigned.
  2. Menin inhibition should not remove H3K79me2 genome-wide. A genome-wide loss would mean menin is required for DOT1L activity generally, which is a different and much stronger claim than the fusion-directed one and would predict marrow toxicity the approved menin inhibitors do not show.
  3. In KMT2A::MLLT10 the two units converge, so a menin inhibitor should behave differently there - either less selectively or more toxically. This is a checkable clinical prediction in a rare but real genotype, and the menin-inhibitor trials enrolled KMT2A-rearranged disease without, as far as this beat's reading goes, reporting outcome by partner at this resolution.
What a refutation would cost the project

This is the section the brief asked for, so it is stated plainly.

The menin-interface hypothesis is refuted, or badly damaged, if:

  • menin inhibition removes H3K79me2 at fusion targets and across the genome in equal measure, because then menin is a general DOT1L cofactor and the therapeutic selectivity has some other source;
  • menin inhibition removes no H3K79me2 at fusion targets while still killing the cells, because then the mark is not on the causal path from the menin interface to cell death and the mechanistic story linking menin, KMT2A and H3K79 methylation is wrong in its middle;
  • the residual fraction is not MLLT10-dependent, because the two-unit model's second unit would then be unidentified;
  • and most damaging of all: if the partition does not exist in KMT2A::AFF1. The CRISPR-ChIP work does not name its fusion in the abstract, and AFF1 is the fusion of infant B-ALL. A menin-dependent oncogenic H3K79 unit demonstrated in an AF9- or ENL-containing system and absent from AFF1 would mean the graph has been importing an AF9 mechanism into its central disease.

None of these is exotic. All four are measurable in existing cell lines with an assay that already exists (ICeChIP), which is why the experiments are specified rather than the hypothesis merely asserted.

System resolved 2026-09-20: the partition is murine MLL-AF9

This hypothesis rests on the CRISPR-ChIP partition of H3K79 into a menin-dependent and a menin-independent unit, and its first open question asked which MLL fusion the study used - the paper does not say clearly.

Answered from the deposited data rather than the text. The GEO series linked from the paper give organism: Mus musculus and cell line: MLLAF9, with sample titles such as MLLAF9 H3K79me2 CRISPR-ChIP day 9 and a menin inhibtor treatment 48hrs arm. The human arms are K562 and profile H3K4me3 and Pol II only.

So the partition is murine MLL-AF9 myeloid, two steps from infant B-lineage KMT2A::AFF1 - different species, different partner, different lineage. That belongs in the claim under rule 73, and this hypothesis should be read as a transfer argument rather than a result in the index disease.

Facts and figures
statementMenin inhibition depletes the fusion-directed oncogenic H3K79 methylation unit and spares the MLLT10-directed endogenous unit, so residual H3K79me2 after menin inhibition is predicted, quantifiable, and should be MLLT10-dependent.
originliterature-gap
prior probability0.55
why this matters to the projectThis is the strongest unbiased functional-genomic support the menin-interface hypothesis has received, and simultaneously the sharpest available test of it.
predictions
[
  "Menin inhibition to full target engagement in a KMT2A::MLLT3 or KMT2A::MLLT1 line leaves a measurable H3K79me2 residue at fusion target loci and leaves genome-wide H3K79me2 largely intact.",
  "MLLT10 knockout collapses that residue.",
  "In KMT2A::MLLT10 the two units are not separable, so menin inhibition there is either less selective or more toxic than in AF9/ENL-partnered disease.",
  "A pharmacodynamic endpoint defined as total H3K79me2 will under-report menin-inhibitor target engagement, because most of the signal is the unit the drug is not meant to remove."
]
experiments
[
  "exp-h3k79-state-resolved-icechip-dose-series",
  "exp-dot1l-degrader-vs-inhibitor-matched-exposure"
]
analyses
[
  "ana-dot1l-dependency-lineage-split-raw-effect"
]
falsifiers
[
  "Menin inhibition removes H3K79me2 genome-wide as well as at fusion targets - menin would then be a general DOT1L cofactor and the selectivity of menin inhibitors has a different source.",
  "Menin inhibition removes no H3K79me2 at fusion targets while still killing the cells - the mark would not be on the causal path from the menin interface to cell death.",
  "The residual H3K79me2 after menin inhibition is not MLLT10-dependent - the second unit would be unidentified and the partition mis-assigned.",
  "The partition cannot be demonstrated in a KMT2A::AFF1 model - the mechanism would then have been established in an AF9/ENL system and imported into infant B-ALL without support."
]
How this could bite us back
  • If menin inhibition spares the endogenous H3K79 unit, then a leukaemia able to re-route its programme onto MLLT10-directed methylation has a resistance mechanism that requires no mutation in MEN1 and would be invisible to any MEN1 sequencing-based resistance assay.
  • Conversely, if a combination is used to remove both units, the endogenous unit is the one normal haematopoiesis runs on, so the combination gives up the selectivity that made the menin inhibitor tolerable.
  • A pharmacodynamic endpoint built on total H3K79me2 measures mostly the unit the drug is not targeting, so a correctly working menin inhibitor could be judged to have failed target engagement.
Still open
  1. RESOLVED 2026-09-20 from the study's own GEO records, not from the paper. elink from PMID 37679565 to GDS returns GSE192562/192561/192555/192544; the per-sample SOFT records give organism 'Mus musculus', cell line 'MLLAF9', and sample titles including 'MLLAF9 H3K79me2 CRISPR-ChIP day 9', 'MLLT10 G1/G2' and treatment 'menin inhibtor treatment 48hrs' [sic]. The human arms are K562 and carry H3K4me3/Pol II only. **So the H3K79 partition this hypothesis rests on is MURINE MLL-AF9, not a human KMT2A::AFF1 system** (rule 73). See `ana-crispr-chip-menin-h3k79-partition-system-resolved-to-murine-mll-af9`.
  2. Has anyone measured H3K79me2 by calibrated ChIP after menin inhibition at full target engagement? A residual fraction is a prediction nobody appears to have reported.
  3. Do the menin-inhibitor trials show a different response pattern in KMT2A::MLLT10, where the two units converge? Outcome by fusion partner at that resolution is not in this graph.
Every source behind this entry (3)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    CRISPR-ChIP, coupling pooled CRISPR screening to chromatin immunoprecipitation on an integrated reporter, reveals a functional partitioning of H3K79 methylation into two distinct regulatory units: an oncogenic DOT1L complex directed by the MLL fusion protein in a menin-dependent manner, and a separate endogenous DOT1L complex whose catalytic activity is directed by MLLT10. The same approach identified all the non-redundant COMPASS members required for H3K4me3 and showed RNA polymerase II is dispensable for maintaining H3K4me3.

    CRISPR-ChIP reveals selective regulation of H3K79me2 by Menin in MLL leukemia. · Gilan O, Talarmain L, Bell CC, et al. · Nat Struct Mol Biol 2023
    PMID 37679565DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. The readout is a synthetic integrated reporter carrying a cis-regulatory element and a single guide RNA, not the endogenous HOXA cluster, so 'directed by the MLL fusion protein' is established on a designed template. **The fusion is now resolved and it is not the index fusion: murine MLL-AF9.** It is not named in the abstract, and the full text is unreachable, but the study's GEO depositions name it per sample. In GSE192555 all eight H3K79me2 ChIP-seq samples are Mus musculus with `cell type: MLLAF9`, including both DOT1L-KO guides, both MLLT10-KO guides and **the menin-inhibitor arm** (GSM5751590). GSE192561 is 24/24 murine. Across all 65 samples in the three series, **none mentions AFF1, AF4 or any AFF-family gene**. The human arm is K562 — a CML line with no KMT2A rearrangement — carrying H3K4me3, Pol II, and five H3K79me2 samples used as a mouse/human library-mix species-spike normalisation under a DOT1L inhibitor, not as a fusion system. So this result reaches KMT2A::AFF1 only by import across fusion, species and lineage at once. Verified by the coordinator directly against the GEO SOFT records on 2026-09-19, which is the single most important missing fact for this hypothesis. A screen identifies requirement, not mechanism, and the authors describe H3K79me2's oncogenic function as 'putative'. **Regraded moderate -> established on 2026-09-20** under `concept-the-confidence-grading-rule`: the functional partitioning of H3K79 methylation is the paper's own headline result, graded established in its sibling entries.

    Evidence that argues against this (2)
    • A menin requirement on a reporter is equally consistent with menin acting upstream of fusion targeting in general: displacing the fusion from chromatin would remove locus-specific H3K79 methylation as a side-effect without menin regulating DOT1L at all. The screen cannot distinguish 'menin is required for the oncogenic DOT1L unit' from 'menin is required for the fusion to be anywhere'. Neville D et al. 2026
    • The fusion behind this result is murine MLL-AF9, not KMT2A::AFF1 — established from the study's own GEO depositions, since the full text is unreachable. Every H3K79me2 sample in GSE192555 is Mus musculus, cell type MLLAF9, and no sample in any of the three series mentions an AFF-family gene. The menin-dependence of the oncogenic unit, which is the part this graph leans on hardest, was measured in mouse myeloid MLL-AF9 cells. KMT2A::AFF1 engages DOT1L through a different architecture — AFF1 carries neither the AHD that MLLT3 and MLLT1 use nor the AF10 OM-LZ — so the partition may not transfer, and `hyp-index-fusion-h3k79-is-entirely-endogenous` is the sharper alternative. NCBI Gene Expression Omnibus 2022
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    AF10 is itself a recurrent fusion partner - in KMT2A::MLLT10 and in PICALM::MLLT10 - and endogenous AF10 regulates progressive H3K79 methylation and HOX expression, with AF10 inactivation impairing transformation by MLL-AF9, MLL-AF6 and NUP98-NSD1.

    AF10 regulates progressive H3K79 methylation and HOX gene expression in diverse AML subtypes. · Deshpande AJ, Deshpande A, Sinha AU, et al. · Cancer Cell 2014
    PMID 25464900DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. Mouse models; the three fusions tested are myeloid and none is KMT2A::AFF1. The relevance to this hypothesis is structural rather than experimental - AF10 being both the endogenous director and a fusion partner is what makes the two-unit model partner-dependent.

    Evidence that argues against this (1)
    • If endogenous AF10 is required for transformation by MLL-AF9, MLL-AF6 and NUP98-NSD1 - fusions that do not contain AF10 - then the 'endogenous' unit is not merely a normal-cell housekeeping activity but is itself part of the oncogenic requirement. That blurs the two-unit partition at its most important boundary. Gilan O et al. 2023
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Menin inhibition induces PRC1.1-dependent deposition of H2AK119ub to silence a subset of MLL-fusion targets, whereas DOT1L inhibition produces a genome-wide increase in H2AK119ub; the PRC1.1 response to DOT1L inhibition arises specifically from progressive loss of H3K79 methylation, independent of MLL-fusion displacement.

    DOT1L provides transcriptional memory through PRC1.1 antagonism. · Neville D, Ferguson DT, Heikamp EB, et al. · Nat Cell Biol 2026
    PMID 41634341DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read; 2026, single report. Cited here because the targeted-versus-genome-wide contrast is exactly what the two-unit model predicts and is the closest thing to independent support the partition has - but the paper does not frame itself as testing the partition and does not measure residual H3K79me2 after menin inhibition.

    Evidence that argues against this (1)
    • The contrast could arise without any partition of H3K79 methylation: menin inhibition acts at fusion-bound loci and DOT1L inhibition acts on a mark present across all transcribed gene bodies, so targeted-versus-genome-wide follows from where each drug's target sits and not from two DOT1L complexes. Gilan O et al. 2023
Connected to